There has been a change to the reports of the top 500 supercomputer list. They now indicate which OSes are run on the systems in the top 500. A ***whopping*** 78.2% (or 391) of all the top 500 supercomputers run a linux flavor or Linux compined with something else. Not sure what that is an indication of. A mixed Linux + ? OS? What's up with that? Oh yeah, 4 BSDs and 5 Macs. One of those MACs is *homegrown* and made the top 15, homegrown being made by the company, which looks like it may do some hush-hush work for the DOD.
Jack wrote:
There has been a change to the reports of the top 500 supercomputer list. They now indicate which OSes are run on the systems in the top 500. A ***whopping*** 78.2% (or 391) of all the top 500 supercomputers run a linux flavor or Linux compined with something else. Not sure what that is an indication of. A mixed Linux
- ? OS? What's up with that? Oh yeah, 4 BSDs and 5
Macs. One of those MACs is *homegrown* and made the top 15, homegrown being made by the company, which looks like it may do some hush-hush work for the DOD.
That wouldent be suprising.. the mac's have a video processing card made specificaly for visualising sonardata and such. I stumbled across that card one day and havent found it since :/ It wouldent suprise me if that was doing image prosessing on satalite data or somthing
I've read things online about Apple employees and/or subcontractors doing work for the Navy, etc. It would make sense for Apple to have the ability to build their own largescale machine to testing Xgrid for various "projects".
On 11/15/05, Kendrick [email protected] wrote:
Jack wrote:
homegrown being made by the company, which
looks like it may do some hush-hush work for the DOD.
That wouldent be suprising.. the mac's have a video processing card made specificaly for visualising sonardata and such. I stumbled across that card one day and havent found it since :/ It wouldent suprise me if that was doing image prosessing on satalite data or somthing
--- Jon Pruente wrote:
I've read things online about Apple employees and/or subcontractors doing work for the Navy, etc. It would make sense for Apple to have the ability to build their own largescale machine to testing Xgrid for various "projects".
I was referring to non-vendor built supercomputers, hence the reason to call them homegrown. Apple does make and market clustered MACs, just like IBM, HP and other manufacturers do. IBM and HP have collectively ~80% of the supercomputer market. Cray is somewhere between 3% and 6%, although per processor Crays seem to outperform all the others. Crays just don't make them as big as IBM and HP. I hear FUD coming out of M$ again about them doing a clustered computing that will be better than the others. Funny how they always come out with this FUD every time there is a new top 500 list. I see no logic in this other than FUD. It's a very small niche market, and will remain so. Of course with the advent of the $100 PC, it might become feasible to expand that market. Anyone with $30,000 could build a top 500 cluster from $100 PCs ... maybe. Realistically, though I don't see anyone building a top 500 for less than $150,000, including the cost of programming and building. Tack on $100,000 - $1,000,000 for M$ licensing and I don't see where they expect to find suckers to buy an M$ Cluster-well-u-know version 2003 (available sometime in 2010). That vaporware stuff is always so nifty. "I don't have a running model, but if I did you'd be like ... oooooh, aaaaaahhh!"
I wonder if we could pool every one on the KCLUG and KULUA UGs and maybe some of the other UGs in the area, how big of a cluster we could throw together, with various mismatched systems. An interesting possibility would be to build a PCMAC cluster. That would be challenging, but probably not very efficient. Although might have some real world uses.
Brian JD
There is a local theoretical physicist I know of that is wanting to build a cluster and probably base it on Linux. I know he's mentioned renting time on the machine to local Univ.s and others to help pay for it. He had some really good local sites reseached to locate it, and his own project to run on it would be quite revolutionary. I'm just hoping that when the build comes on I can get my hand dirty with setting up some highend hardware.
Jon
On 11/16/05, Jack [email protected] wrote:
I wonder if we could pool every one on the KCLUG and KULUA UGs and maybe some of the other UGs in the area, how big of a cluster we could throw together, with various mismatched systems. An interesting possibility would be to build a PCMAC cluster. That would be challenging, but probably not very efficient. Although might have some real world uses.
Brian JD
Jon Pruente wrote:
There is a local theoretical physicist I know of that is wanting to build a cluster and probably base it on Linux. I know he's mentioned renting time on the machine to local Univ.s and others to help pay for it. He had some really good local sites reseached to locate it, and his own project to run on it would be quite revolutionary. I'm just hoping that when the build comes on I can get my hand dirty with setting up some highend hardware.
Jon
On 11/16/05, *Jack* <[email protected] mailto:[email protected]> wrote:
I wonder if we could pool every one on the KCLUG and KULUA UGs and maybe some of the other UGs in the area, how big of a cluster we could throw together, with various mismatched systems. An interesting possibility would be to build a PCMAC cluster. That would be challenging, but probably not very efficient. Although might have some real world uses. Brian JD
Im also intrested in seeing what the via c7 and dual core systems are like for clustering. C7 is suposed to be much faster and intended for laptops. If the correct board layout was made they could be extreemly eficent for clustering.. L.J had a article about a 16 node cluster in a tool box. they have made several. one used p3 mini-itx but needed a huge power supply. the pc104 units are near 100 the via's are 140 ish+ so it culd be done for relitivly inexpensivly right now.
-
Jack -
Jon Pruente -
Kendrick